Demand driven deep brain stimulation: Regimes and autoregressive hidden Markov implementation

Deep brain stimulation is an increasingly prevalent surgical option in the treatment of a multitude of neurological conditions, most notably Parkinson´s disease. The development of a neurofeedback device is driven primarily by stimulator habituation, surgical risk factors, the cost of battery replacement, and reported neuropsychiatric side-effects under prolonged chronic administration. Here we present two distinct regimes for stimulation delivery in chronic and acute symptomatic conditions, presented in the context of Parkinsonian bradykinesias and tremor. Implementation strategies are discussed with a focus on vector-autoregressive hidden Markov models for tremor prediction. Detection of simple motor actions versus tremor are compared in a preliminary performance analysis.